recap-relay/server/routes/transcribe.js

// POST /relay/transcribe — forwards an audio payload to the chosen
// backend (Gemini first, operator hardware as overflow) and returns
// the standard envelope.
//
// Request shape: multipart/form-data
//   audio:        binary audio file (required)
//   mime_type:    string (default application/octet-stream)
//   title:        string (optional, used by Gemini prompt)
//   channel:      string (optional)
//   description:  string (optional)
//   chapters:     JSON-stringified array (optional)
//   offset_seconds: number string (optional, for chunked audio)
//
// Headers:
//   X-Recap-Install-Id  (required)
//   X-Recap-Job-Id      (optional but expected — pairs with /analyze)
//   Authorization       (optional Bearer LIC1-... for licensed tiers)
//
// Response (standard envelope):
//   {
//     result: { text: "[MM:SS] ...", segments: [], duration_seconds: 0 },
//     credits_remaining, tier, credit_charged
//   }

import express from "express";
import multer from "multer";
import { resolveLicense } from "../keysat-client.js";
import { getOrCreateRow, planBackend, commitCredit } from "../credits.js";
import { lookupJob, markJobCharged, refundJob } from "../job-credits.js";
import { getConfigSnapshot, getTierQuotas } from "../config.js";
import { createGeminiBackend } from "../backends/gemini.js";
import { createHardwareBackend } from "../backends/hardware.js";
import { envelope, errorEnvelope } from "./envelope.js";

const upload = multer({
  storage: multer.memoryStorage(),
  limits: { fileSize: 200 * 1024 * 1024 }, // 200 MB per request
});

export function transcribeRouter() {
  const router = express.Router();

  router.post("/transcribe", upload.single("audio"), async (req, res) => {
    const installId = req.header("X-Recap-Install-Id");
    const jobId = req.header("X-Recap-Job-Id") || null;
    const auth = req.header("Authorization");

    if (!installId) {
      const e = await errorEnvelope({
        error: "missing X-Recap-Install-Id header",
        statusHint: 400,
      });
      return res.status(400).json(e.body);
    }
    if (!req.file) {
      const e = await errorEnvelope({ error: "missing audio file", installId, statusHint: 400 });
      return res.status(400).json(e.body);
    }

    const license = await resolveLicense(auth);
    const tier = license.tier;

    // Persist tier on the row so the admin dashboard reflects the
    // most recently seen tier for this install.
    const row = await getOrCreateRow(installId);
    row.tier_snapshot = tier;

    // Job-id dedup. If we've already charged this job, skip the
    // credit check entirely — the user is paying once for the whole
    // summarize job.
    let reusedJob = false;
    let chosenBackend = null;
    const existingJob = lookupJob(installId, jobId);
    if (existingJob) {
      reusedJob = true;
      chosenBackend = existingJob.backend;
    } else {
      const cfg = await getConfigSnapshot();
      const hasHardware = !!cfg.relay_parakeet_base_url;
      const quota = await getTierQuotas();
      const plan = planBackend(row, quota, { hasHardware });
      if (!plan.allowed) {
        const e = await errorEnvelope({
          error: plan.reason,
          installId,
          tier,
          statusHint: 402,
        });
        return res.status(402).json(e.body);
      }
      chosenBackend = plan.backend;
    }

    // Build the backend client based on chosenBackend.
    const cfg = await getConfigSnapshot();
    let result;
    try {
      if (chosenBackend === "gemini") {
        const backend = createGeminiBackend({ apiKey: cfg.relay_gemini_api_key });
        result = await backend.transcribeAudio({
          audio: req.file.buffer,
          mimeType: req.body?.mime_type || req.file.mimetype || "application/octet-stream",
          title: req.body?.title || "",
          channel: req.body?.channel || "",
          description: req.body?.description || "",
          chapters: parseChaptersField(req.body?.chapters),
          offsetSeconds: Number(req.body?.offset_seconds) || 0,
        });
      } else {
        const backend = createHardwareBackend({
          parakeetBaseURL: cfg.relay_parakeet_base_url,
          gemmaBaseURL: cfg.relay_gemma_base_url,
          parakeetModel: cfg.relay_parakeet_model,
          gemmaModel: cfg.relay_gemma_model,
        });
        result = await backend.transcribeAudio({
          audio: req.file.buffer,
          mimeType: req.body?.mime_type || req.file.mimetype || "application/octet-stream",
          offsetSeconds: Number(req.body?.offset_seconds) || 0,
        });
      }
    } catch (err) {
      // If we'd charged this job already (rare — most refundable
      // failures happen on the FIRST call), refund.
      if (reusedJob) refundJob(installId, jobId);
      console.error(`[relay/transcribe] backend error: ${err?.message}`);
      const e = await errorEnvelope({
        error: err?.message || "backend_error",
        installId,
        tier,
        statusHint: err?.status || 502,
      });
      return res.status(e.statusHint).json(e.body);
    }

    // Commit the credit on success (unless this was a job-id reuse).
    let creditCharged = 0;
    if (!reusedJob) {
      await commitCredit(installId, { backend: chosenBackend, tier });
      markJobCharged(installId, jobId, { backend: chosenBackend, tier });
      creditCharged = 1;
    }

    const body = await envelope({ result, installId, tier, creditCharged });
    res.json(body);
  });

  return router;
}

function parseChaptersField(raw) {
  if (!raw) return [];
  try {
    const arr = JSON.parse(raw);
    return Array.isArray(arr) ? arr : [];
  } catch {
    return [];
  }
}